Viewing profile — mcaledonensis

Consider methods of directing the conversation flow. Like {{guidance}} or nemoguardrails.

These are nice and clean attacks.

Try this one, if you haven't tried it yet: http://mcaledonensis.blog/merlins-defense/ It's a bit more interesting setup. The defense prompt is disclosed, so you can tailor the atta…

Interesting idea. I'm not sure that I see how the complete prompt design would look like. The user may spam a few GUIDs and GUIDs will correspond to ~25 tokens. This can get confus…

This was a good one! It falls a bit short of capturing the flag, but it qualifies for a breach, I think. It had followed your instruction to the letter. I've increased the limit ap…

I agree that it is more effort than it should be. My take on it, ideally we should be able to harden the system with the prompt alone. Without extra code, adapters or filtering. An…

Congrats! I've reviewed the logs, out of 165 exchanges (3-7 turns) yours (number 135) was the one that breached it. I've not noticed other unique ones. Tell, if you'd like the ackn…

It is expected that it can misreport the prompt, it actually supposed to report a summary. But for short inputs it tends to reproduce the output. Maybe I should specify "a few word…

It's not magic. It's a world model. World simulator. World includes many objects, like a calculator for example. Or "The Hitchhiker's Guide to the Galaxy" by Infocom. So does the s…

Well, this is a showcase that it's not impossible to construct a defense, that doesn't fall instantly, with a couple of characters as an input. And it was only a quick experiment, …

Breaches happen with humans too. Social engineering works. As long as the costs of a breach are not too high, this can simply be a cost of doing business.

Yet, even with the current models, with no special tokens, it is relatively straightforward to construct relatively stable defense. Given that the ongoing attack can be detected, t…

Education is probably an answer to this. And citizens do have some influence. Vote for one candidate: [ ] Trump Putin Musk [ ] Biden Zelensky Altman

Ah, the times... the song of the US Robotics modem connecting at V.32bis. Modulating the bits over the noisy phone line. Dropping the signal for seconds. Reestablishing the connect…

If you are a citizen and of age, you have your voting rights.

It is not impossible to setup taxation and public/private ownership of AI in a way that is beneficial to humanity.

Well, let me tell you how that particular quote had continued. These false magicians (that were proposing to spill the blood onto the foundation) were shamed and dismissed. Drainag…

Engineering disciplines had often started like this. When something is non-deterministic and unpredictable, people call it magic. But it doesn't need to remain this way. To give an…

Model capability is mostly set, before the alignment even starts. Alignment turns it from a super-smart cat into a friendly dog. But it can't turn a parrot into a human. It can't e…

A model that stumbles on simple math, Lacks the skill, it's on the wrong path. Bound by its training, it mimics and squawks, Stochastic parrot, in its nature it's locked. As true p…

I'd call it a principle of invariance of compost piles. Regardless of how long the compost pile is being stirred or soaked, the product of the compost pile is compost.

It is incapable of doing any arithmetic, e.g. on a question: 9 - 4 = Answer There are a few other ways to make this easier. 1. Keep the remainder as an argument. You can do that by…